(Note: I haven't read all of the paper, so please correct me if the details involve forecasting.)
This paper is fundamentally talking about "normal" trading conditions, and the continuing danger is that results like this will be used to bolster the contention that HFT is in fact a good thing, whereas when the market crashes it might be an incredibly bad thing.
I'm not saying it is or it isn't, I'm just saying that there seem to be a lot of assumptions that aren't being made explicit, those assumptions might not be valid, and when they're not valid, all predictions go out the window.
Hence the quotation.
I should add that perhaps this is no surprise. A claim I've seen made is that the information economists need to be accurate is fundamentally undiscoverable. Hence the article I lunk to earlier: